Item/Trigger for Detecting Crashed Machines

[Nate Bell]

Hi,
This may be a simple question, but I haven't been able to find an answer:
We have a computer that has been crashing recently, and I want Zabbix to send an alert to several people when this machine crashes. Obviously I need a Trigger that will toggle when Zabbix cannot reach the computer in question, and the Trigger must be linked to an Item monitoring some status on the computer.

Here's my problem: everything I have tried to monitor stops monitoring when the computer crashes. Any zabbix_agentd dependant item will obviously be unable to report values, but as far as I know, internal, simple, and SNMP checks should keep reporting even if they can't find the host they are looking for. If this is true, I can't seem to get any data at all from these once a host has turned off. I'm thinking I don't have SNMP set up correctly.

The {{HOSTNAME}:tick.nodata(180)}=1 trick mentioned in the ZABBIX manual seems like it would work, but it says it needs data from zabbix_sender, and I can't find any information on using zabbix_sender.

So, I suppose my question is: what's the best way to monitor a host to see if it has crashed, and set up a trigger to activate when a host crashes.

Thanks,
Nate

[Nate Bell]

Solved

After a lot of poking around, I found I needed to install fping to get icmpping to work correctly, which does the task I wanted it to. Now I have an item that changes values when a machine reboots/crashes so I can apply a trigger to it to monitor for crashes.

Here is the thread that helped solve my problem, if anyone is interested.

Nate

[bbrendon]

How would this work for active agents that are not pingable?

[Alexei]

ICMP pings do not require any agent running on hosts being pinged. It's like a simple ping command running on ZABBIX server machine.

[hairball]

Hi Alexei,

Reading this thread, I get the impression I'm better off using icmp ping rather than the status flag for monitoring hosts availability then? In the documentation it suggests:

"Two methods (or combination of both methods) may be used in order to monitor availability of a server.
ICMP ping (Key "icmpping")
Key "status"

But the following trigger using status to monitor if any of 5 hosts goes down seems to fail "Error: Trigger just added. No status update so far." in that the availability graphs show no information (unknown: 100%).

({server1.somewhere:status.last(0)}=2) | ({server2.somewhere:status.last(0)}=2) | ({server3.somewhere:status.last(0)}=2) | ({server4.somewhere:status.last(0)}=2) | ({server5.somewhere:status.last(0)}=2)

zabbix is gathering other information from these hosts just fine.

I was wondering am I doing this the fundamentally the wrong way or have I just made a minor error?

Thanks in advance

[James Wells]

Greetings,

Reading this thread, I get the impression I'm better off using icmp ping rather than the status flag for monitoring hosts availability then? In the documentation it suggests:

As stated in the documentation, status is a good way to test. Status takes the icmmping or agent ping. As such, it can be used in more situations that just icmpping. Personally, I use status to create events / alerts instead of pings. Seems to be more reliable to me.

But the following trigger using status to monitor if any of 5 hosts goes down seems to fail "Error: Trigger just added. No status update so far." in that the availability graphs show no information (unknown: 100%).

That is correct. And you have just pinged on the only real issue with using status. Because status is a calculated value, rather than a tested value, it starts out undefined, and remains that way until the ping values that it calculates changes state.